Magic’s Post

July 2022: Russia had invaded Ukraine, the market had just crashed to its absolute low (the NASDAQ had crashed by 30%, Bitcoin by 70%) - in an age before ChatGPT and generative AI - and with little hope for the tech world, a very young and ambitious, very curly-haired gentleman came to us through a mutual friend, Daniel. He had a vision that stood out bright against the grey mood at the time, fittingly called Magic. His eyes flashed when he described a monumental vision: an AI developer you can talk with and who truly understands code. The vision seemed crazy ambitious, but why did we even believe him he could have a real shot at realizing it? Today, I want to share what fascinated me about Eric's founder journey in our first conversation and what concerns he had to overcome: Eric Steinberger's journey from a curious teenager to an influential figure in AI is a tale of exploring technology and pursuing bold visions. From his early days at a technical high school in Vienna, where he and his friend Sebastian, at age 11, practiced their coding skills by hacking their school's computer lab, Eric's path has been anything but ordinary. Eric's academic pursuits took him to University of Cambridge to study computer science at a young age. He was discovered when his implementation and enhancement of a paper from Meta Facebook AI Research Lab (FAIR) caught the attention of the AI community. He became its youngest employee as a working student and his contributions to a reinforcement learning paper that set new standards for efficiency in learning from limited data further established his reputation, all while being a working student. Never let anyone tell you you are too young! Despite an offer to join a tempting full time position at Facebook AI Lab, Eric chose a different path. He left Cambridge and FAIR behind to address a pressing global issue: climate change. Founding ClimateScience, he grew the NGO to become a leading force in climate education, powered by over a thousand volunteers worldwide. Eric's journey came full circle when he co-founded Magic. It's this personal story of pushing the limits that for me makes the difference in supporting a vision that you would not think anyone else can accomplish. In supporting Eric and Magic, I've learned the importance of pushing beyond initial doubts and recognizing the potential in young, visionary leaders. Their age and the audacity of their dreams are not limitations but rather indicators of their capacity to think differently and drive significant change in society. Here's to more visionaries like Eric, who inspire us to challenge our hesitations and invest in a future where technological science fiction becomes reality.

A glimpse of where the future of #GenerativeAI driven #software #engineering is heading. Seems somewhat inline with what Jensen Huang, the CEO of NVIDIA said a few days back (https://lnkd.in/d5fSfKYu) One of the coolest use cases of GenAI I have seen till date. Blog: https://lnkd.in/dZc4wrSE #generatieveai #artificialintelligence #softwaredevelopment #softwareengineering #technology #coding

Vercel ai SDK with object stream is 🤌

Magic Secures $117M for AI Software Engineering Advancements Magic, a company focused on code generation, secures $117M in funding led by notable investors like Nat Friedman and Daniel Gross. The investment aims to advance AI software engineering, emphasizing code generation as a pathway to Artificial General Intelligence (AGI). This endeavor requires pioneering algorithms, extensive CUDA utilization, large-scale training, reinforcement learning (RL), and innovative user interfaces. The funding enables Magic to expand its capabilities in AI model development. The company invites talented individuals to join its mission of solving complex challenges and contributing to the development of safe AGI, leveraging thousands of GPUs in a collaborative environment. #Magic #AI #SoftwareEngineering #CodeGeneration #NatFriedman #DanielGross #Funding #ArtificialGeneralIntelligence #AGI #CUDA #ReinforcementLearning #RL #UserInterfaces #ModelDevelopment #GPU #Collaboration #Innovation

Excited to share my latest project showcasing a face recognition model built with OpenCV, face recognition, and NumPy. 🚀 Using these tools, the model adeptly detects and recognizes faces – think Elon Musk, Bill Gates, and Steve Jobs! Leveraging OpenCV's powerful functionalities alongside the face recognition library, this project highlights the seamless identification and labeling of these prominent figures. Stay tuned for more exciting updates! #machinelearning

Oso's declarative permissions language, Polar, is not Turing complete. That's a good thing because you don't want your permissions language to potentially run forever. For user-facing permissions apps, you need your permissions queries to be low latency. Turing completeness would allow for our users to shoot themselves in the foot, performance-wise, in ways we could neither automatically detect nor mitigate. Read our proof on why Polar isn't Turing complete and why that's a good thing

1) Helped build a CNN to detect emotional valence which was frustrating while building by rewarding when it worked. So far I have basically only worked with CSVs with pandas, working with other types of input like audio is definitely more challenging. 2) Working with people forced me to use git properly and illustrated how important having a clean workflow was. Something I did not have to think about when working alone. 3) If I am having to use Collab for compute on my tiny project I wonder how any AI startup is going to compete with Big Tech - probably on speed of development/iterations. I knew GPUs were important just not to what extent.

The second cohort of our “From users to builders — AI-powered productivity for tech roles” is about to reach full enrollment — we just opened the third cohort. Many people complained that it’s too far away. Sorry. To make a good course in our spare time, we have to take longer. But we have prepared a free 30 minutes version on April 21th. You can sign up here: https://lnkd.in/gjTmFUa7 Although it’s free, it’s really a distilled version of our full course: through building a prototype, we introduce the new paradigm of accessing compute — from GUI to AI — and the mindset behind it.

My recent switch from Colab Pro to Lightning.ai was one of the best decision I have made for my data science experiments! 🚀 Lightning.ai offers unmatched flexibility in coding and integrates seamlessly with the awesome VSCode editor. Tools like Chainlit, Streamlit, TensorBoard, and MLflow, which were a bit challenging on Colab, now run effortlessly on Lightning.ai. Plus, it provides persistent storage for my workspace, so no more reloading data or setting up the environment every time! The best part? You get 24/7 access to a studio with a 4-core CPU for free! 🌟 #DataScience #LightningAI #VSCode #MachineLearning #AI

Another great course from #DeepLearning.AI I highly recommend this course to all software engineers.